Abstract IA026: International multicentrer validation and clinical implementation of AI-driven support for ultrasound detection of ovarian cancer

Background: A critical shortage of expert ultrasound examiners has raised concerns of unnecessary interventions and delayed cancer diagnoses. Artificial intelligence (AI)-driven diagnostic support has the potential to alleviate this burden and improve patient outcomes. Deep learning, applied to ultrasound images, has recently demonstrated promising results in ovarian cancer detection; however, external validation is lacking. Our aim was to develop a deep learning model for ovarian cancer detection and evaluate its robustness and ability to generalize in a large multicenter setting. We also sought to assess an AI-assisted triage strategy and compare it to current practice in a retrospective simulation. Methods: We retrospectively collected 17,119 ultrasound images from 3,652 women with an ovarian lesion (2,224 benign, 1,428 malignant) from 20 centers in 8 countries. A total of 2,718 cases were externally reviewed by a minimum of 7 experts and 6 non-expert examiners. Using a leave-one-center-out cross-validation scheme, for each center in turn, we trained a transformer-based model using data from the remaining centers and compared the models' performance with that of expert and non-expert examiners in terms of accuracy, sensitivity, specificity, and F1 score. Furthermore, we simulated and assessed how these models could be used in AI-assisted triage and compared it to current practice. Findings: The models demonstrated robust performance across centers, ultrasound systems, and histological diagnoses, with an overall area under the receiver operating characteristic curve (AUC) of 0.93 (95% CI, 0.92–0.94), and F1 score of 83.5% (95% CI, 81.8–85.1) on cases from unseen centers. They outperformed both expert and non-expert examiners, with F1 scores of 79.5% (95% CI, 77.6–81.2, p < 0.0001]) and 74.1% (95% CI, 72.1–76.1, p < 0.0001]), respectively. The models were further shown to produce well-calibrated predictions. In a retrospective simulation, AI-assisted diagnostic support reduced the number of referrals to experts by 63%, from 52% of cases (current practice) to 19%, while increasing the diagnostic performance (F1 77.2% vs 82.7%, p < 0.0001]). Interpretation: The models exhibit strong generalization and outperform expert examiners. Introducing AI-driven diagnostic support into the clinical workflow may reduce human resource demands, while improving diagnostic performance. Citation Format: Elisabeth Epstein. International multicentrer validation and clinical implementation of AI-driven support for ultrasound detection of ovarian cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Ovarian Cancer Research; 2025 Sep 19-21; Denver, CO. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl):Abstract nr IA026.